The Developmental Origins of Dehumanisation

The Developmental Origins of Dehumanisation

Niamh Caitriona McLoughlin

PhD

University of York

Psychology

2

Abstract

Dehumanisation is a pervasive social phenomenon that has facilitated historical and modern examples of extreme violence, prejudice and discrimination. The perception that a person can be ‘less human’ than another person is typically applied to social outgroup members who are attributed with fewer uniquely human capacities compared to ingroup members. A significant amount of developmental research has examined the origins of intergroup bias among young children, however, investigation into the development of our tendency to dehumanise others has been relatively neglected. This is despite the fact that dehumanisation is closely linked to children’s social cognitive understanding (e.g., mental state inference) and behaviour (e.g., prosociality). The aim of my doctoral studies was to investigate the developmental origins of this phenomenon. The results of the empirical work in Chapters 2 and 3 revealed that 6-year-olds perceive outgroup faces to be physically less human than ingroup faces and that even younger children (5-year-olds) are less likely to reference the mental states of individuals belonging to a different group. The final experimental chapter (Chapter 4) explored the effects of encouraging children to mentalise about the behaviour of a perceived outgroup and showed that this technique is sufficient to increase empathic helping towards an outgroup member in need. The implications of this research for the nature of dehumanisation in development, as well as for children’s understanding of human and non-human agents, are discussed. Ultimately, further inquiry into how dehumanising biases emerge, and are potentially learnt, could contribute to strategies focused on improving intergroup relations.

3

List of Contents

Primary Supervisor Statement... 11

Secondary Supervisor Statement ... 12

Chapter 1: General Thesis Introduction ... 13

Dehumanisation... 14

Early Psychological Perspectives ... 14

Infrahumanisation ... 16

The Dual Model of Humanness ... 17

The Stereotype Content Model ... 18

Mind Perception ... 19

Summary ... 20

The Development of Social Cognition ... 21

Agency Understanding ... 21

Mental State Understanding ... 22

Summary ... 25

The Development of Intergroup Cognition ... 26

Explicit and Implicit Preference... 26

Social Essentialism ... 30

4

Summary ... 31

The Development of Dehumanisation ... 32

The Present Research Aims ... 34

Chapter 2: Children’s Perception of Humanness ... 36

Introduction ... 37

Study 1 ... 41

Method ... 42

Results and Discussion ... 48

Study 2 ... 49

Method ... 50

Results and Discussion ... 53

General Discussion... 56

Chapter 3: Biases in Children’s Mental State Attribution ... 60

Introduction ... 61

Method ... 62

Results ... 68

Discussion ... 72

Chapter 4: Improving Children’s Intergroup Relations ... 76

Introduction ... 77

Study 1 ... 81

Method ... 81

Results and Discussion ... 86

Study 2 ... 87

Method ... 88

5

General Discussion... 92

Chapter 5: General Thesis Discussion ... 95

The Nature and Trajectory of Dehumanisation in Development ... 96

Defining Dehumanisation ... 96

Dehumanisation and Preference ... 98

Dehumanisation and Essentialism ... 100

Dehumanisation and Status ... 101

Dehumanisation in Different Intergroup Contexts ... 102

Summary and Final Thoughts ... 104

Theory of Mind Research... 105

Children’s Perception of Non-Human Agents ... 106

The Cultural Transmission of Dehumanisation ... 108

Reducing Prejudice ... 109

Conclusion ... 110

6

List of Figures

Page

Figure 1. The face morph stimuli (4 female, 4 male) used in the test trials for Studies

1 and 2 (taken from Hackel, Looser, & Van Bavel, 2014). ... 43

Figure 2. The materials used in the training phase of the studies including the

response scale (panel A), the four liquid measurements (panel B) and the four points along one of the training morph continua (panel C) that corresponded to the four points along the humanness scale. ... 47

Figure 3. The results from Study 1 with gender groups (panel A) and Study 2 with

geographically based groups (panel B) when age was treated as a categorical variable (error bars represent one standard error). ... 54

Figure 4. The relationship between age (in months) and relative dehumanisation

from Study 1 with gender groups (panel A) and Study 2 with geographically based groups (panel B). ... 55

Figure 5. The results for the mean number of mental state words (panel A) and for

the mean diversity of mental states words (panel B) that 5- and 6-year-old children produced in each condition. Error bars represent the standard error of the mean. .... 72

Figure 6. Examples of the storybook images (created at www.storyboardthat.com)

7

Figure 7. Materials: (A) the four-point scale that measured children’s ratings of

emotion intensity, (B) the cardboard trays that were part of the helping task (note that the tray belonging to the participant was placed nearer to them while the one belonging to the victim was placed slightly further away) and (C) the ‘ruined’

drawing used as a prop for the emotion intensity measure in Study 1. ... 83

Figure 8. The results from Study 1 (panel A) and Study 2 (panel B) for the mean

number of stickers that children donated to the victim. Error bars represent standard error of the mean. ... 93

8

Acknowledgements

To Harriet, someone mentioned to me at the beginning of my PhD that you are a rare combination of ‘brilliance and warmth’ and I have come to wholeheartedly agree with this sentiment (and not just because you are an ingroup member - social psych joke!). I am deeply grateful to you for inspiring me to be as ambitious as possible throughout the last few years, always with a warm smile and a constant string of encouragement. I attribute my continuing passion for finding out what children think about the world to you, thank you. To Steve and Janine, thank you for readily imparting your expertise at any time and for your kind and reassuring words.

To my York family - you know who you are - I did not realise completing a doctorate would end up being some of the most hilarious and enjoyable years of my life, I owe that to you guys. To my York Irish family, the craic at training and during the aul seisiúns has been a massive help in pushing me through this final year, go raibh maith agaibh agus York Eireannach abú!

To Giacomo, I do not know how I would have coped without coming home to your patient hugs and warm bowls of pasta, grazie mille amore. To my home girls, you are forever my rocks. To Seán and Róisín, even though we live in different countries for most of the year, thank you for making your entertaining and comforting presence felt.

And now, the big one, to Mam and Dad, any of my past, present and future accomplishments are and will always be a direct reflection of your unconditional love and support.

9

Author’s Declaration

I, Niamh McLoughlin, declare that this thesis is a presentation of original work and I am the sole author. This work has not previously been presented for an award at this, or any other, University. All sources are acknowledged as References.

The empirical work presented in this thesis has been published or is currently under review in the following peer-reviewed journals:

McLoughlin, N., Tipper, S. P., & Over, H. (2017). Young children perceive less humanness in outgroup faces. Developmental Science. Advanced online publication. doi:10.1111/desc.12539.

All authors contributed to the design of this study. C. Looser and T. Wheatley granted the first author permission to use and publish a sample of their stimuli. N. McLoughlin collected and analysed the data under the supervision of H. Over and S. P. Tipper. A. Eggleston performed reliability coding. N. McLoughlin drafted the manuscript and H. Over made a number of revisions. All authors approved the final manuscript before submitting it for publication.

McLoughlin, N., & Over, H. (2017). Young children are more likely to spontaneously attribute mental states to members of their own group. Psychological

Science. Advanced online publication. doi:10.1177/0956797617710724

Both authors contributed to the design of this study. U. Frith and colleagues granted the first author permission to use their stimuli. N. McLoughlin collected and

10

analysed the data under the supervision of H. Over. S. P. Tipper provided some valuable advice regarding the coding and analysis of the data. A. Eggleston performed reliability coding. N. McLoughlin drafted the manuscript and H. Over made a number of revisions.

McLoughlin, N., & Over, H. (2017). Encouraging young children to mentalise about

a perceived outgroup increases empathic helping towards them. Manuscript

submitted for publication. [Currently under review in Developmental Science]

Both authors contributed to the design of this study. The picture book stimuli were created using Storyboard That online software. N. McLoughlin collected and analysed the data under the supervision of H. Over. A. Eggleston performed reliability coding. N. McLoughlin drafted the manuscript and H. Over made a number of revisions. M. Carpenter provided some valuable comments on an earlier draft of the manuscript.

11

Primary Supervisor Statement

I am listed as a co-author on the three empirical papers which make-up the main body of this thesis.

In each of the reported studies, the work is primarily that of Ms. Niamh McLoughlin. For each paper, Niamh compiled the relevant stimuli, completed all of the data collection and coded and analysed the summary data. Niamh wrote the first draft of each paper.

(See hard copy for signature)

12

Secondary Supervisor Statement

I am listed as second co-author on the empirical paper which makes-up Chapter 2 of this thesis.

This work is primarily that of Ms. Niamh McLoughlin. Niamh compiled the relevant stimuli, completed all of the data collection and coded and analysed the summary data. Niamh wrote the first draft of this paper.

13

Chapter 1: General Thesis Introduction

The ability to recognise and to reason about the minds of our fellow human beings is essential to social functioning (Baillargeon et al., 2013; Dennett, 1996; Gray, Gray, & Wegner, 2007; Tomasello, 1995; Woodward, Sommerville, & Guajardo, 2001). These skills appear early in development, such that even young infants display sensitivity to the social and mental agency of others (Carpenter, Nagell, & Tomasello, 1998; Johnson, 2000; Johnson, 2003). Yet, despite this initial understanding, social psychologists have demonstrated that there are situations in which adults deny the mental life, and in general the humanity, of other people. For example, outgroup members are typically perceived as having fewer uniquely human traits and mental state faculties (e.g., intelligence, culture, complex emotional experiences) compared to ingroup members (Harris & Fiske, 2006; Haslam & Loughnan, 2014; Leyens et al., 2001). This phenomenon is known as ‘dehumanisation’. Dehumanisation lies at the root of harmful social problems ranging from genocidal atrocities and extreme intergroup violence (Bandura, Barbaranelli, Caprara, & Pastorelli, 1996; Chalk & Jonassohn, 1990; Leidner, Castano, Zaiser, & Giner-Sorolla, 2010; Tirrell, 2012) to more everyday examples of prejudice and discrimination (Bastian & Haslam, 2010; Čehajić, Brown, & González, 2009). Given these implications for the development of social cognition and behaviour, it is quite surprising that the psychological origins of this phenomenon have received very little research attention to date.

The purpose of my doctoral studies is to investigate dehumanising biases in young children. I aim to contribute to the small body of work on this topic and to provide a clearer picture of when children first begin to selectively attribute

14 humanness to social group members. In addition, I wish to explore the potential link between children’s mental state understanding and dehumanisation by assessing whether they are more likely to spontaneously refer to the mind of ingroup members relative to outgroup members. I ultimately hope that this work can inform interventions to combat the negative consequences of this phenomenon and to encourage positive group relations.

Dehumanisation

Before delving into the relevant developmental literature, it is important that I first outline how dehumanisation has been defined in psychological research with adults. Although this phenomenon is observed in other contexts, for example, in the objectification of females in the media (Loughnan et al., 2010; Vaes, Paladino, & Puvia, 2011), the treatment of medical patients (Haque & Waytz, 2012; Vaes & Muratore, 2013) and even in the perception of the self (Bastian et al., 2013; Haslam, Bain, Douge, Lee, & Bastian, 2005), it is generally thought of as the denial of humanness to social outgroup members relative to ingroup members (Allport, 1954; Leyens, Demoulin, Vaes, Gaunt, & Paladino, 2007). Psychological theories tend to converge on the intergroup nature of dehumanisation; however, they differ somewhat with regards to the key features of this bias.

Early Psychological Perspectives

Social psychologists were originally interested in dehumanisation as a psychological mechanism to explain mass violence carried out in times of conflict (Bar-Tal, 1989; Schwartz & Struch, 1989; Staub, 1989; Tajfel, 1981). The majority of theorists focused on how the blatant act of dehumanising a person, or an entire group, weakens moral inhibitions (Bandura, 1991, 1999; Kelman, 1973) and thus

15 excludes them from moral concern (Opotow, 1990), allowing for their cruel treatment with little to no distress on part of the perpetrators. Some prime examples of this process include the likening of Jews to vermin in Nazi propaganda campaigns (Bytwerk, 1983), the comparisons between the Tutsi social class and ‘cockroaches’ in the lead up to the Rwandan genocide (Kellow & Steeves, 1998; Tirrell, 2012) and the categorisation of Black people as ‘savages’ during the slave trade in the colonial period (Jahoda, 1999). In these and similar cases, brutal actions were deemed to be justifiable because the eradication and/or oppression of marginalised groups were beneficial for society (Smith, 2012).

The first experimental study illustrating the behavioural effects of dehumanisation was carried out in support of Albert Bandura’s (1991, 1999) theory of moral disengagement. Bandura, Underwood, and Fromson (1975) showed that participants who heard a dehumanising description of a novel group (e.g., ‘animalistic’ and ‘rotten’) punished this group more for making mistakes (through the apparent administration of electric shocks) than when the recipients were characterised in more human terms (e.g., ‘perceptive’ and ‘understanding’).

In contrast to these explicit, absolute denials of humanness, more recent psychological theories have paid attention to the implicit aspects of dehumanisation in society more broadly. The most influential of these accounts take an attribute-based approach, as opposed to the metaphor-attribute-based approach in which outgroup members are directly compared to non-human categories (see above and Goff, Eberhardt, Williams, & Jackson, 2008). For this perspective, empirical studies investigate and identify the traits considered to be uniquely human and then ask adults to distribute these characteristics among social groups.

16

Infrahumanisation

Leyens and colleagues (2000, 2001) were the first to operationalise the more subtle and relative way we ascribe humanness to others in the absence of intergroup tension. They assume that we reserve the full human ‘essence’ for ingroup members in the form of traits that distinguish humans and animals, namely intelligence, language, reasoning and emotions. This assumption derives from essentialist beliefs about social group categories in which we attribute members of a certain category with an underlying quality that determines their nature (Gelman, 2003; Hirschfeld, 1995). Therefore, outgroup members are perceived to be inherently less human or to possess an ‘infra-human’ essence. Empirically, these theorists concentrated on the distribution of a type of mental state, i.e., basic (primary) and uniquely human (secondary) emotions. They showed that basic emotions (e.g., sadness, happiness, fear) - thought to be biologically-driven and shared with animals (Ekman, 1992) - are differentiated from uniquely human emotions (e.g., remorse, hope, nostalgia) - which represent a more complex depiction of human interaction (Kemper, 1987) - in lay conceptions of humanness (Demoulin et al., 2004).

Subsequent studies revealed that ingroup members are judged to experience significantly more positive and negative secondary emotions compared to primary emotions (Boccato, Cortes, Demoulin, & Leyens, 2007; Paladino et al., 2002). In comparison, there are no significant differences in the type of emotions attributed to members of a perceived outgroup. This effect was coined outgroup ‘infrahumanisation’ and has been connected to an array of problematic behavioural outcomes, largely outside the domain of conflict and violence (Vaes, Paladino, Castelli, Leyens, & Giovanazzi, 2003), such as less empathy for and less willingness to help outgroup members in crisis (Čehajić et al., 2009; Cuddy, Rock, & Norton,

17 2007), fewer instances of intergroup forgiveness (Tam et al., 2007) and reduced perspective taking (Vaes, Paladino, & Leyens, 2004). Furthermore, researchers have identified potential moderators of this bias. Increased support for nationalistic and conservative ideals (DeLuca-McLean & Castano, 2009; Viki & Calitri, 2008), as well as less knowledge of, low perceived similarity and less contact with outgroups (Brown, Eller, Leeds, & Stace, 2007; Rodríguez Pérez, Delgado Rodríguez, Betancor Rodríguez, Leyens, & Vaes, 2011), predict greater levels of infrahumanisation.

The Dual Model of Humanness

Haslam (2006) developed a theory that also focuses on how adults characterise humanity and assign these traits to themselves and to other people (Haslam et al., 2005). However, he proposes that we conceptualise humanness along two dimensions: one based on the comparison between human and animals (uniquely

human; similar to infrahumanisation) and another based on the comparison between

humans and automata (human nature; Loughnan & Haslam, 2007). The denial of essential, uniquely human traits (e.g., rationality, culture) is termed ‘animalistic dehumanisation’ and associated with feelings of disgust. The denial of core, human nature traits (e.g., individuality, interpersonal warmth) is termed ‘mechanistic dehumanisation’ and generally promotes psychological distance. Empirical findings have reflected these two senses of humanness - the attribution of uniquely human and human nature qualities are seen to be uncorrelated (Haslam et al., 2005) and this distinction is observed across cultures (Bain, Vaes, Kashima, Haslam, & Guan, 2011; Martinez, Rodriguez-Bailon, & Moya, 2012). Interestingly, adults attribute more positive and negative uniquely human traits (e.g., polite, disorganised) to ingroup members than to outgroup members but only think of their ingroup as higher in negative human nature terms (e.g. insecure, jealous; Koval, Laham, Haslam,

18 Bastian, & Whelan, 2011). The authors argue this finding represents a group protective function whereby emphasising the flaws of ingroup members make them ‘only human’. This is an interesting interpretation but further study is needed to explore why valence seems to matter in the attribution of qualities related to mechanistic dehumanisation.

Unlike Leyens (2001) infrahumanisation model (principally related to less severe outcomes for intergroup behaviour), Haslam’s (2006) theory proposes a continuum of behavioural consequences that range from very dangerous tendencies to more everyday discrimination. For example, adults who were exposed to low uniquely human and human nature descriptors of Muslims were more likely to report a necessity to torture Muslim prisoners of war (Viki, Osgood, & Phillips, 2013) and quicker associations between female images and both animal- and object-related terms predicted men’s proclivity to sexually harass women (Rudman & Mescher, 2012).. In addition, both animalistic and mechanistic dehumanisation have been correlated with lower levels of outgroup helping (Andrighetto, Baldissarri, Lattanzio, Loughnan, & Volpato, 2014), and outgroup praise (Bastian, Laham, Wilson, Haslam, & Koval, 2011).

The Stereotype Content Model

The infrahumanisation (Leyens et al., 2000, 2001) and the dual model (Haslam, 2006) theories of dehumanisation imply that the attribution of fewer mental states to outgroup members is an important feature of this harmful phenomenon. In their research on the neural bases of dehumanisation, Harris and Fiske (2009, 2011) exclusively define this bias as a failure to consider the mind of another person. Their work stems from the stereotype content model (Fiske, Cuddy, Glick, & Xu, 2002)

19 within which group members are evaluated on their perceived warmth and

competence. Ingroup members are judged to be both likeable and highly competent

whereas members of ostracised outgroups (e.g., homeless people, drug addicts) are generally disliked and deemed lower in educational and/or economic status. Hence, these outgroups are likely to be dehumanised.

Potentially supporting this idea, neuroscientific data have demonstrated that adult participants display less activity in brain regions associated with ‘mentalising’ (i.e., medial prefrontal cortex; Frith & Frith, 2006) and increased activation in disgust-related areas (e.g., insula; Wicker et al., 2003) when viewing extreme outgroup members. Moreover, they used fewer mental state verbs when asked to describe a day in the life of a, for example, homeless person (Harris & Fiske, 2006). Although the authors have not directly associated differences in brain activity with adverse intergroup behaviour, Harris and Fiske (2011) suggest that this cognitive failure could contribute to the inhumane treatment of others (also see Fiske, 2009).

Mind Perception

Very closely related to this approach, another body of research argues that how we perceive the minds of other people is inextricably linked to the concept of dehumanisation (Waytz, Epley, & Cacioppo, 2010; Waytz, Gray, Epley, & Wegner, 2010). The two methods through which we can deny the mind, or ‘dementalise’, others are very similar to Haslam’s two aspects of humanness (Haslam & Loughnan, 2014; Waytz et al., 2010). Specifically, Gray and colleagues (2007) found that the perception of agency, or mental state abilities that distinguish humans from animals (e.g., to plan, to think) and experience, or emotional capacities that distinguish humans from inanimate entities (e.g., to feel pain, to have a personality) are involved

20 in mind detection. These theorists further elucidate the relation between mind perception and dehumanisation when considering the impact of attributing someone with less of a mind on moral decision-making. For instance, if we do not perceive another person as having a mind that experiences distress, then we will not proceed to empathise with their distress (Bruneau, Cikara, & Saxe, 2015) and to include them within the realm of moral concern (Waytz et al., 2010).

Empirically, adults perceive a liked target to be more mindful compared to a disliked target (Kozak, Marsh, & Wegner, 2006). More related to intergroup bias, Hackel, Looser, and Van Bavel (2014) suggest that ingroup members are attributed with mental states more readily than are outgroup members. Across two studies, they presented participants with a series of doll-human face morphs ranging from 0% animate to 100% animate and manipulated their belief about whether the morphs were based on members of their own group or another group. The results revealed that ingroup faces were perceived to have a mind at a lower threshold on the continuum (approx. 60%) than outgroup faces (approx. 70%).This effect was later replicated with participants from both Eastern and Western cultures (Krumhuber, Swiderska, Tsankova, Kamble, & Kappas, 2015).

Summary

Taken together, the definition of dehumanisation in social psychological research seems to be a complex and multifaceted one. However, there are a few critical features that overlap across different accounts. The denial of mental state and other human-like characteristics to outgroup members and outgroup derogation (dislike and perceived lower status) are commonly associated with dehumanising perceptions. Reasoning about social groups in an essentialist manner also seems to

21 be an important component in the process of dehumanisation. These social and psychological factors have also been studied in a distinct body of literature investigating the development of social and intergroup cognition in children.

The Development of Social Cognition

Some of the definitions outlined above suggest that, in order to dehumanise, one must first understand what ‘humanness’ entails to some extent (Bain et al., 2011; Castano et al., 2009; Demoulin et al., 2004). A significant research effort in developmental psychology has been dedicated to identifying when the recognition and understanding of human action and thought emerges in development (Baron-Cohen, 2000; Cutting & Dunn, 1999; Leslie, 1994; Wellman & Bartsch, 1988) and what this means for children’s social (Chalik, Rivera, & Rhodes, 2014; Jenkins & Astington, 2000) and cognitive (Carlson, Moses, & Breton, 2002; Ozonoff, Pennington, & Rogers, 1991) competence.

Agency Understanding

Researchers interested in the perception of agency have pinpointed certain features that may help infants to recognise potential interaction partners: whether an agent has a face or eyes, whether it engages in a contingent, social interaction and whether it appears to move in a self-propelled way (Beier & Carey, 2014; Johnson, 2000; Premack, 1990). For example, infants follow the gaze of a motorised object significantly more often (Johnson, Slaughter, & Carey, 1998) and make social evaluations about the apparent behaviour of moving shapes (Hamlin & Wynn, 2011; Hamlin, Wynn, & Bloom, 2007; Powell & Spelke, 2013) when they possess some of these human-like properties.

22 Developmental research has often relied on these anthropomorphic tendencies to shed light on the development of children’s social cognitive abilities. Five-year-olds, but not younger children, perceive social intent in the seemingly purposeful interaction between animated shapes (see Heider & Simmel, 1944, for comparable work with adults) and are able to identify which shape is ‘scared’ or ‘being mean’ (Springer, Meier, & Berry, 1996). Similar stimuli were used to examine mentalising deficits in autism and found that autistic children produced fewer and less appropriate mental state descriptions of these animations compared to matched controls (Abell, Happé, & Frith, 2000). Research into when children first begin to mentalise about the behaviour of others has been especially prolific within the developmental literature and could have particularly interesting implications for the concept of dehumanisation.

Mental State Understanding

Our ‘theory of mind’ refers to the capacity to overcome an egocentric way of thinking and to reason about the mental states (i.e., intentions, desires, beliefs, emotions) of other people (Premack & Woodruff, 1978; Wellman, 1990). Previous research has shown that young infants demonstrate a form of intention understanding; for example, 7-month-olds are more surprised by collision events that involve two moving people compared to two moving objects (i.e., a box and cylinder mounted on wheels), perhaps expecting that humans would try to avoid colliding with one another (Woodward, Phillips, & Spelke, 1993). Other research has looked at infants’ understanding of goal-directed actions in more depth (Woodward, 1998). Five-, 6-, and 9-month-olds were habituated to an actor reaching for one of two toys presented at a distance. At test, the position of these toys was switched and the actor reached for the same toy (but with a different motion pattern) in one trial and for the

23 alternate toy (but with the same motion pattern as before) in another trial. The results revealed that the older infants looked longer during trials where the actor chose a different toy (and not when they moved in a different direction). Importantly, looking times were not affected when a metal rod engaged in the same movement patterns during test trials. A similar experiment with 18-month-olds found that they were able to correctly interpret an adult’s intended action through observation of their failed attempts (Meltzoff, 1995) but that they did not produce the target acts when the disrupted actions were performed by a set of mechanical arms. Thus, infants seem to infer that human motion, and not object motion, is driven by internal goals and intentions (Woodward, Phillips, & Spelke, 1995).

Further work has suggested that older infants can discriminate between intentional and accidental actions (Carpenter, Akhtar, & Tomasello, 1998) and can predict someone’s desired food choice based on their prior reactions (Repacholi & Gopnik, 1997). The exact emergence of belief understanding is in contention however (Heyes, 2014; Wellman, Cross, & Watson, 2001). Traditional tests have focused on children’s ability to indicate the false belief (i.e., a belief about the world that differs from the reality of the situation) of another individual (Wimmer & Perner, 1983). For example, in the famous Sally Anne task (Baron-Cohen, Leslie, & Frith, 1985), children view a puppet putting an object into one of two locations (e.g., box A) and then observe the experimenter changing the location of the object (e.g., by putting it in box B) in the puppet’s ‘absence’. Children’s task is to specify the location where they think the puppet would look for the object when it returns (i.e., box A). The results of this paradigm have demonstrated that typically developing children are reliably able to adopt the mental perspective of this agent from around the age of 4. However, more recent work employing less demanding, nonverbal

24 versions of this test (e.g., based on looking time, active helping measures; Buttelmann, Carpenter, & Tomasello, 2009; Onishi & Baillargeon, 2005; Southgate, Senju, & Csibra, 2007) have argued that young infants demonstrate this ability.

Another aspect to this line of work is examining how individual differences in theory of mind proficiency are associated with children’s socio-cognitive skills. As alluded to above, a persistent difficulty in understanding the mental states of others is considered a central characteristic in autism spectrum disorder (ASD) and leads to low quality and confusing social interactions for these individuals (Baron-Cohen, 2000; Chevallier, Kohls, Troiani, Brodkin, & Schultz, 2012). For typically developing children, their performance on theory of mind assessments predicts the nature of peer play (Nielsen & Dissanayake, 2000), how they evaluate moral actions (Baird & Astington, 2004) and their level of executive function (Carlson et al., 2002) among other social cognitive outcomes (Moore & Frye, 1991). Other research with slightly older children has shown that second-order false belief understanding (i.e., Person X wrongly believes that Person Y believes the object is in location A; Perner & Wimmer, 1985) plays a role in coordinating with peers (Grueneisen, Wyman, & Tomasello, 2015) and in the correct interpretation of situational evidence(Astington, Pelletier, & Homer, 2002). Furthermore, their attribution of second-order ignorance (i.e., Person A does not know what Person B knows) helps them to distinguish between a lie and a joke (Sullivan, Winner, & Hopfield, 1995).

Emotion understanding

Since a person’s mental states also encompass his or her emotional experiences (Hadwin & Perner, 1991; Hughes & Dunn, 1998) and emotion perception is a common measure in dehumanisation research (Leyens et al., 2000;

25 Paladino et al., 2002), it is worthwhile to mention when the understanding of primary and secondary emotions may be observed in development. Past work has suggested that the comprehension of basic emotions appear early in life (Sroufe, 1983) - very young infants discriminate between happy and sad vocal and/or facial expressions (Field & Walden, 1982; Vaillant-Molina, Bahrick, & Flom, 2013) and attend more to the gaze of fearful faces, than to neutral or happy faces, when learning about novel objects (Hoehl, Wiese, & Striano, 2008). From around the age of 5 or 6, children can make inferences about complex socially-driven emotions like remorse and guilt (Vaish, Carpenter, & Tomasello, 2011) and are able to verbalise their experience of embarrassment (Bennett, 1989; Buss, Iscoe, & Buss, 1979) and pride (Seidner, Stipek, & Feshbach, 1988) to some degree. Children’s understanding of secondary emotions is thought to continually develop at a significant rate throughout middle childhood (Harris, Olthof, Terwogt, & Hardman, 1987; Tangney & Fischer, 1995).

Summary

It is clear that, from early in development, children are sensitive to and are able to reason about various features of humanity. Developmental research has concentrated on when these capacities emerge (Baillargeon et al., 2013; Flavell, Green, & Flavell, 1990; Johnson, 2000), as well as causes and consequences of these skills (Astington, 2001; Heyes, 2014). What are less clear are the social circumstances in which children may deny the humanness of other people.

The Development of Intergroup Cognition

Investigation into the origins of intergroup prejudice has provided empirical evidence to suggest that young children might perceive outgroup members as less

26 human than ingroup members. More specifically, these studies have shown that children demonstrate some of the social biases previously associated with dehumanisation in adults, such as strong preferences for their own versus other groups (Dunham & Emory, 2014), essentialist beliefs about social group members (Rhodes, Leslie, & Tworek, 2012) and knowledge of the relative status of social classes (Liben, Bigler, & Krogh, 2001; Olson, Shutts, Kinzler, & Weisman, 2012).

Explicit and Implicit Preference

The psychological work examining the development of intergroup bias has primarily focused on children’s relative implicit and explicit preferences for group members (Dunham & Degner, 2010). It has revealed that, within Western societies at least, social categories based on gender, language and racial differences are particularly influential in the formation of children’s attitudes (Aboud, 1988; Hilliard & Liben, 2010; Kinzler, Dupoux, & Spelke, 2007; La Freniere, Strayer, & Gauthier, 1984). In addition to these types of group membership, I will also discuss social divisions based on place of origin or nationality because they are highly relevant in the current social and political climate (Sanneh, 2016; Schmuck & Matthes, 2015) and therefore in my empirical chapters.

Gender

Children profess explicit liking for same-gender individuals from at least 2 years of age (Martin & Ruble, 2004; Yee & Brown, 1994). This preference extends into how they reason about and engage with their social environment. In a set of experiments, Shutts, Banaji, and Spelke (2010) found that 3-year olds base their preferences for novel objects and activities on the gender (but not the race) of the endorser. Additionally, 3-year olds attend to the gender (but, again, not the race) of

27 other children when deciding who they should be friends with and who would share their social preferences (Shutts, Roben, & Spelke, 2013). Young children even demonstrate this positivity in an implicit way by responding faster to associations between photos of same-gender peers and positive words in a child-friendly version of the Implicit Association Task (Dunham, Baron, & Banaji, 2015).

Language

Preferences for members of our own language group also emerge very early in development (Kinzler et al., 2007). Five- and 6-month-old infants preferentially attend to native language speakers, and older infants (both American and French) are more likely to accept toys from individuals who speak their own language. Relatedly, Buttelmann, Zmyj, Daum, and Carpenter (2013) observed that 14-month olds imitate the actions of a native speaker more faithfully than actions performed by a foreign language speaker. Five-year-old American children favour peers who speak in their native language or with their native accent compared to peers who speak in a foreign accent (i.e., English in a French accent; Kinzler & Spelke, 2011). These preferences are judged to be robust and can even sometimes override young children’s liking for other social groups – they prefer to affiliate with racial outgroup members who spoke in a native accent compared to the reverse (Kinzler, Shutts, DeJesus, & Spelke, 2009).

Race

Research has shown that 9-month-old infants orient towards own-race faces (Kelly et al., 2005) and by 3- to 4-years of age, young children can categorise people by race (Dunham, Stepanova, Dotsch, & Todorov, 2015). Even though Shutts et al. (2013) did not find that race influenced 3-year-olds’ social choices, 4-year-olds use

28 race to infer third-party relationships. From around the age of 5, children express social preference for members of their own racial group (Doyle & Aboud, 1995; Kinzler et al., 2009; Kircher & Furby, 1971) and 6-year-old children assign more positive traits to their racial ingroup at an implicit level (Baron & Banaji, 2006; Dunham, Baron, & Banaji, 2006). However, it is important to note that this developmental trend is mainly true for majority (largely White) race children; the emergence of ingroup preferences in children belonging to minority racial groups is not as straightforward (Margie, Killen, Sinno, & McGlothlin, 2005; Milner, 1973) and considerably influenced by cultural factors (for e.g., see Shutts, Kinzler, Katz, Tredoux, & Spelke, 2011).

Nationality

Nationality is a more complex social grouping because it can involve identifying with one or a culmination of social markers (e.g., national colours, cultural traditions, native accent/language, race; Barrett, 2007). For example, relevant research has illustrated how the recognition of this social category is a particularly dynamic process. When presented with the choice, both American and Korean 5- and 6-year-old children tended to categorise national ingroup members based on the language they spoke, regardless of the speaker’s race (DeJesus, Hwang, Dautel, & Kinzler, 2017a). Older children, in comparison, had a broader view of their national ingroup and their consideration of racial and linguistic cues differed between the two cultures. Therefore, it is not surprising that, unlike most of the other group divisions mentioned in this section, emergence of biased attitudes favouring one’s own national group varies significantly across sociocultural contexts (Bennett et al., 2004; Oppenheimer & Barrett, 2011). Children as young as 3 years of age can show explicit negativity towards national outgroups in countries where intergroup

29 conflict is prevalent (Bar-Tal, 1996; Povrzanović, 1997) compared to children living in more peaceful settings. Children in these contexts usually begin to reliably prefer and identify with their own country at about 6 or 7 years of age (Barrett, 2007).

Ingroup positivity versus outgroup negativity

Young children demonstrate robust preferences for ingroup members across multiple group dimensions, yet, this display of ingroup positivity may not be equated to having a negative view of the outgroup. There is evidence to suggest that ingroup positivity does appear somewhat earlier in development than outgroup derogation (Aboud, 2003; Benozio & Diesendruck, 2015; Brewer, 1999). Buttelmann and Böhm (2014) showed the potentially diverging development of ingroup love and outgroup hate in a minimal group paradigm. This type of group manipulation - based on relatively minimal symbols of similarity (for e.g., the yellow vs. the green group)- is valuable for investigating the effects of mere categorisation on young children’s evaluative responses (Dunham, Baron, & Carey, 2011; Richter, Over, & Dunham, 2016) and behaviour (Engelmann, Over, Herrmann, & Tomasello, 2013; Oostenbroek & Over, 2015; Plötner, Over, Carpenter, & Tomasello, 2015). Their findings revealed that 6-year-olds prevented their minimal ingroup from receiving negative items (e.g., a spider) and instead opted to distribute these items to a neutral party. Older children (i.e., 8-year-olds), however, gave significantly more of these undesirable items to the minimal outgroup.

Researchers have provided various explanations for the possibly distinct developmental trajectories of these processes. Some have emphasised the evolutionary value of early attachment to the ingroup (Brewer & Caporael, 2006; Caporael, 1997) while others have focused on the role of learning in the emergence

30 of negative outgroup attitudes (Over, Eggleston, Bell, & Dunham, 2017). Meltzoff (2007, 2013) presents a theoretical perspective that highlights our early propensity to prefer people who are similar to us. Based on research with infants, he argues that early shared experiences with others (e.g., imitating a parent’s actions or expressions and vice versa) shapes ‘like me’ preferences which underlie the ingroup positivity observed in young children.

Social Essentialism

Essentialism was first investigated as an element of children’s folk psychological theories (Gelman & Legare, 2011; Hirschfeld, 1995). This research demonstrated an almost universal tendency to believe that the biological and social world is made up of ‘natural kinds’ that possess a fundamental hidden property which is determined from birth and is responsible for their observable traits (Gelman, 2004). The grouping of biological entities into different kinds is considered helpful for children’s learning. For example, knowing that one tiger is aggressive allows children to make similar inferences about seemingly harmless baby tigers (Waxman, Medin, & Ross, 2007) and those who have a slightly different appearance (Gelman & Legare, 2011) because they all retain a tiger ‘essence’. However, these essentialist beliefs become problematic when they are applied to social groups (Hirschfeld, 1996; Rhodes et al., 2012) and can facilitate stereotyping (Allport, 1954; Haslam, Rothschild, & Ernst, 2000; Prentice & Miller, 2007), and perhaps even dehumanising biases (Leyens, 2009) in adults.

More recent work has examined the developmental trajectory of social essentialism. In these studies, children make judgements about the generalisability of stereotypic characteristics, for example, whether every boy likes to play sports, and

31 if these traits are hereditary and stable, for example, whether the race of a baby will always match the race of the parent. By the age of 5, children are thought to essentialise about gender (Taylor, Rhodes, & Gelman, 2009) and native language (Kinzler & Dautel, 2012) and the emergence of these beliefs seem to be prevalent across cultures (Rhodes & Gelman, 2009). In contrast, the development and strength of essentialist beliefs about racial (Astuti, Solomon, & Carey, 2004; Waxman, 2010), ethnic (Birnbaum, Deeb, Segall, Ben-Eliyahu, & Diesendruck, 2010; Diesendruck & HaLevi, 2006) and religious (Chalik, Leslie, & Rhodes, 2017) groups relies on cultural influences. For instance, young children from more politically conservative backgrounds in the US are more likely to display a racial bias (Rhodes & Gelman, 2009) and Israeli children in religious communities are more likely to use inductive reasoning when talking about Jew and Arab social categories compared to secular children (Diesendruck & Haber, 2009).

Perceived Status

The perception of dehumanised outgroups often co-occurs alongside the belief that outgroup members are inferior to the ingroup in some respect and/or constitute a lower social class (Capozza, Andrighetto, Di Bernardo, & Falvo, 2011; Harris & Fiske, 2006; Loughnan, Haslam, Sutton, & Spencer, 2014). Developmental work has shown that sensitivity to the hierarchical nature of society emerges early (Bigler, Brown, & Markell, 2001; Nesdale & Flesser, 2001). Three-year-olds associate a greater amount of wealth with higher status racial groups in certain cultures (Olson et al., 2012) and older children think novel occupations are more prestigious (e.g., more important and gain a higher wage) when they are portrayed by conventionally competent characters (i.e., White males; Bigler, Averhart, & Liben, 2003; Liben et al., 2001). Moreover, across a variety of contexts, children tend to

32 favour members belonging to high status groups (Bigler et al., 2001; Horwitz, Shutts, & Olson, 2014; Newheiser & Olson, 2012; Shutts et al., 2011).

Summary

Overall, there is strong empirical evidence to suggest that young children are able to understand some of the important components of humanity and, in addition, exhibit many of the intergroup biases linked to outgroup dehumanisation in previous literature. I will now review the few recent studies that have begun to investigate this phenomenon in adolescents and children. Similar to work with adults (Haslam, 2006; Leyens et al., 2001), this small body of work has focused on the attribution of human characteristics to social in- and outgroup members.

The Development of Dehumanisation

Two different studies showed that adolescents, aged between 11 and 16 years of age, ‘infra-humanise’ outgroup members by specifying fewer uniquely human emotions (e.g., sympathy, hope) in the evaluation of peers who belong to a different school (Brown et al., 2007) and who support a different football team (Chas, Betancor, Rodríguez-Pérez, & Delgado, 2015) to them. Martin, Bennett, and Murray (2008) used an adapted version of this paradigm to examine the development of infrahumanisation in 6- to 7-year-old and 10- to 11-year-old Scottish children. In this study, children of both ages believed their national football team would experience secondary emotions (e.g., pride) more intensely than primary (e.g., anger) emotions but that the English football team would experience both types of emotion to a similar degree. Costello and Hodson (2014) also proposed that children judge outgroup members to be less human than ingroup members on this dimension across racial divisions. Their results revealed that 6- to 10-year-old White participants

33 thought Black children experience fewer complex emotions (e.g., guilt) and possess fewer uniquely human capacities in general (e.g., curiosity). Finally, Van Noorden, Haselager, Cillessen, and Bukowski (2014) developed a measure to assess both animalistic and mechanistic forms of dehumanisation towards friends and non-friends in 7- to 12-year-olds. They found that ingroup peers were attributed with more human nature (e.g., sociability, friendliness) and uniquely human (e.g., humility, politeness) qualities and that outgroup peers were more likely to be dehumanised in an animalistic (i.e., denied uniquely human traits) rather than in a mechanistic fashion. Taken as a whole, this research implies children as young as 6 years of age might dehumanise peer, racial and national outgroup members.

Although these trait-based results are compelling, they are limited to some respect in explaining how dehumanising biases originate in development. It is not clear whether adults and young children conceive of humanity in the exact same way; for example, Betancor Rodriguez, Chas Villar, Rodriguez-Perez, and Delgado Rodriguez (2016) recently proposed that 11- and 12-year-olds categorise the respective humanness of emotional terms differently to adults. One of these differences was that, while adults believe secondary emotions to be a strong indicator of a person’s morality, children endorse basic emotions as signs of ‘goodness’ or ‘badness’. Betancor Rodriguez et al. (2016) claim this particular deviation could be reflective of children’s developing sense of morality. Hence, previous findings on this topic should perhaps be interpreted with caution because they may not directly reflect children’s dehumanising perceptions. Furthermore, investigation into the developmental trajectory of dehumanisation and how it relates to other facets of children’s social and intergroup cognition has been somewhat neglected.

34

The Present Research Aims

The first aim of my doctoral work was to take a step away from paradigms based on emotion and trait attribution and to develop a novel measure of humanness for use with young children. My first experimental chapter (Chapter 2) thus explores whether children perceive outgroup members to be less human than ingroup members in the context of face perception. In two studies, 5- and 6-year-olds were asked to rate the physical humanness of ambiguous doll-human faces (taken from Hackel et al., 2014) when they belonged to their gender in- and outgroup (Study 1) and to a geographically based in- and outgroup (Study 2). Potential age differences in this tendency and whether it was associated with children’s explicit preference for the social groups was also investigated.

My next empirical chapter (Chapter 3) combines adult research on dehumanisation with developmental work on mental state understanding to investigate whether young children are biased in the way they mentalise about the behaviour of other people. I chose to focus on this specific feature of dehumanisation because potential biases in children’s mental state attribution could have direct implications for the nature of their theory of mind abilities. For this study, 5- and 6-year-olds were asked to describe the actions of interacting geometric shapes (employed in research with autism; see Abell et al., 2000) and their belief about the group membership of these characters was manipulated (again, across gender and geographically based categories). The dependent measure was the quantity and diversity of mental state content in children’s descriptions.

My final set of studies (Chapter 4) attempts to address the ultimate aim of the present research. Building on the results from Chapter 3, I examined the role of

35 mental state attribution in fostering positive intergroup behaviour (specifically for outcomes that are associated with dehumanisation) in a politically relevant context. To accomplish this, 5- and 6-year-olds were either encouraged to discuss the thoughts and feelings or the actions of an immigrant group and their perception of a novel group member’s emotions was measured, as well as the extent to which they helped another group member in an empathic-based situation.

I decided to recruit 5- and 6-year-old children in my empirical research for multiple reasons. First, by this age, children can explicitly reason about the social behaviour of agents (Springer et al., 1996) and pass standard theory of mind tests (Baron-Cohen et al., 1985; Wellman, 2002). Second, and most importantly for Chapter 3, children in this age range correctly use mental state terms when referring to people’s minds (Shatz, Wellman, & Silber, 1983) and frequently do so in conversation (Frith & Frith, 2003). Finally, 5- and 6–year-olds exhibit the intergroup biases previously linked to dehumanisation: they demonstrate implicit and explicit preferences for members of their own groups (Banaji, Baron, Dunham, & Olson, 2008; Kinzler et al., 2009; Patterson & Bigler, 2006), essentialise about certain social categories (Birnbaum et al., 2010; Kinzler & Dautel, 2012; Rhodes & Gelman, 2009) and are sensitive to the status of social group members (Nesdale & Flesser, 2001).

36

Chapter 2: Children’s Perception of Humanness

Reference:

McLoughlin, N., Tipper, S. P., & Over, H. (2017). Young children perceive less

humanness in outgroup faces. Developmental Science. doi:10.1111/desc.125391

Abstract

We investigated when young children first dehumanise outgroups. Across two studies, 5- and 6-year-olds were asked to rate how human they thought a set of ambiguous doll-human face morphs were. We manipulated whether these faces belonged to their gender in- or gender outgroup (Study 1) and to a geographically based in- or outgroup (Study 2). In both studies, the tendency to perceive outgroup faces as less human relative to ingroup faces increased with age. Explicit ingroup preference, in contrast, was present even in the youngest children and remained stable across age. These results demonstrate that children dehumanise outgroup members from relatively early in development and suggest that the tendency to do so may be partially distinguishable from intergroup preference. This research has important implications for our understanding of children’s perception of humanness and the origins of intergroup bias.

1

The author, Niamh McLoughlin, designed the experiment, collected the data, analysed the results, and wrote the article under the supervision of Dr. Harriet Over and Prof. Steven Tipper.

37

Introduction

In order to navigate the social world, it is essential to be able to recognise and engage with potential interaction partners (Baillargeon et al., 2013; Over, 2016; Tomasello, 1995). A great deal of developmental research has focused on when children are first able to identify socially relevant agents (Johnson, 2000; Meltzoff, 1995; Woodward et al., 1993) and attribute human-like capacities to those agents (Carpenter et al., 1998; Onishi & Baillargeon, 2005; Repacholi & Gopnik, 1997). However, a body of work from social psychology suggests that we do not always consider the humanity of others. Adults tend to ‘dehumanise’, or deny full humanness to, outgroups (Bandura, 1991, 1999; Loughnan et al., 2014; Vaes, Leyens, Paladino, & Miranda, 2012; Viki & Calitri, 2008). Outgroup members are perceived to have fewer uniquely human qualities, such as rationality, openness and cultured beliefs (Haslam, 2006), and are also attributed with fewer second-order emotions, such as compassion and remorse (Leyens et al., 2000; Leyens et al., 2001) than ingroup members. Leyens and colleagues found that this differential attribution of emotion was present across a series of studies and termed the effect ‘infrahumanisation’. More generally, outgroup members are thought to have less of a mind (Hackel et al., 2014; Harris & Fiske, 2006, 2011; Krumhuber et al., 2015). Dehumanisation has been linked to acts of prejudice (Haslam & Loughnan, 2014; Rudman & Mescher, 2012; Viki, Osgood, & Phillips, 2013) and neglect (Čehajić et al., 2009; Cuddy et al., 2007) observed among social groups. The developmental origins of this phenomenon are thus important for our understanding of intergroup relations.

To date, there has been relatively little work on dehumanisation in development. Certainly, we know from previous research that intergroup biases are

38 present from a young age (Aboud, 1988; Dunham et al., 2011; Kinzler et al., 2009; Patterson & Bigler, 2006). Research has shown that young children reliably exhibit both implicit and explicit preference for individuals of the same gender (Dunham et al., 2015; Yee & Brown, 1994). For example, Shutts et al. (2010) found that 3-year-olds prefer novel objects and activities that are endorsed by same-gender peers. Gender is also the first social category that children think about in an essentialised manner, in that they believe group members share an underlying quality or ‘essence’ that defines their nature (Gelman, 2003; Rhodes & Gelman, 2009). From around the age of 5, children prefer and assign more positive traits to their own racial group (Doyle & Aboud, 1995; Kinzler & Spelke, 2011). Somewhat older children (i.e. 6-year-olds) show implicit own-race preference, at a level comparable to that seen in adults, as measured by a child friendly version of the Implicit Association Task (Baron & Banaji, 2006). With regards to national groups, children begin to explicitly identify with and prefer their own country from around the age of 6 or 7 (Barrett, 2007). The emergence of these attitudes, however, varies across sociocultural settings and can be seen earlier in countries that have recently experienced or are currently experiencing intergroup conflict (Oppenheimer & Hakvoort, 2003; Teichman, 2001).

Only a handful of studies have considered the origins of dehumanisation. These studies have tended to focus on emotion and trait attribution. Van Noorden et al. (2014) asked 7- to 12-year-old children to judge if their friends versus their non-friends possessed human-like qualities such as humility, trustworthiness and sociability. In general, children thought of friends as having more of these traits. A related finding looking specifically at social groups found that White children, aged 6 to 10 years, attributed fewer human traits (e.g., curiosity, creativity) and fewer

39 second-order emotions (e.g., embarrassment, love, guilt) to Black targets than to White targets (Costello & Hodson, 2014). Finally, Martin et al. (2008) showed how 6- to 11-year-old Scottish children estimated that their national football team would experience second-order emotions (e.g., pride, disappointment) more intensely than would the English football team. Although these results regarding trait and emotion attribution are suggestive, they need to be complemented by more extensive research investigating which qualities young children actually associate with humanness (for e.g., see Betancor Rodriguez et al., 2016, with older children).

To our knowledge, no studies have yet explored the developmental origins of dehumanisation in relation to social perception. We therefore examined whether young children perceive outgroup members to be less human. To investigate this question, we focused on face perception since previous research has demonstrated that young children are able to perceive social qualities in faces. Cogsdill, Todorov, Spelke, and Banaji (2014) found that 5- and 6-year-olds were at adult levels of reliability when judging faces for trustworthiness, dominance and competence and, in addition, Song, Over, and Carpenter (2016) found that similarly aged children are able to discriminate between subtly different facial expressions (i.e., real vs. fake smiles). Other research has suggested that group membership influences how young children perceive faces. At least from the age of 5, and perhaps even considerably younger (Bar-Haim, Ziv, Lamy, & Hodes, 2006), children are better at recognising the faces of ingroup members, including same-race (Feinman & Entwisle, 1976; Pezdek, Blandon-Gitlin, & Moore, 2003) and same-age (Anastasi & Rhodes, 2005) individuals. In terms of social group bias, 3- and 4-year-old White children are more likely to categorise racially ambiguous faces as outgroup members when they possess a negative expression (Dunham, Chen, & Banaji, 2013).

40 We investigated whether children perceive less humanness in outgroup faces relative to ingroup faces. In order to do this, we adapted a paradigm from the adult literature to make it suitable for developmental research. Hackel et al. (2014) presented a set of face stimuli that were generated by morphing doll faces with human faces to create a series of continua that ranged from 0% animate (i.e., doll face) to 100% animate (i.e., human face). In two studies, Hackel et al. (2014) manipulated the group to which these faces belonged by informing participants that some of the faces were based on morphs developed from ingroup members and others were based on morphs developed from outgroup members. Participants were then asked to rate the extent to which each face looked like it ‘had a mind’ on a 7-point scale. Results indicated that the threshold for perceiving a mind in a face was lower for ingroup members, when fewer human cues were present, (at approx. 60% increment along the continuum) compared to the threshold for outgroup faces (at approx. 70% increment along the continuum). In other words, ingroup faces were humanised, and perceived to have a mind, more readily than were outgroup faces.

We modified this paradigm in the following ways. First, we substantially reduced the number of trials by identifying the most ambiguous doll-human morph from each face continuum in a pretest study with adults. This allowed us to have eight test trials rather than the 110 that were presented to adult participants. Second, we modified the test question and, instead, asked participants how human the face appeared. Our final modification involved the way in which children gave their responses. We asked them to estimate how human each face looked on a 4-point scale ranging from ‘not at all human’ to ‘completely human’. We predicted that children would judge the morph faces to be less human when they belonged to their outgroup than when they belonged to their ingroup.

41 We chose to examine this question with 5- and 6-year-old children because we know that they are able to extract social meaning from faces (Cogsdill et al., 2014), that social categories influence how they process faces (Dunham et al., 2013; Pezdek et al., 2003) and that they show reliable preferences for their own groups (Dunham & Emory, 2014; Kinzler et al., 2007; La Freniere et al., 1984). Lastly, by including 5-year-olds, we tested dehumanisation in a somewhat younger age group than other research on this topic.

Study 1

We tested the extent to which children perceive relatively less humanness when evaluating other-gender faces. We chose gender because it is a particularly salient category for young children when thinking about and engaging with their social environment (Grace, David, & Ryan, 2008; Martin & Ruble, 2004; Shutts et al., 2013). In an informal pilot study with 32 5- to 7-year-olds (18 boys, Mage = 6;5, age range = 5;7-7;5), we presented the ambiguous doll-human faces selected in the pretest study with adults (see Figure 1) and found that children, on average, perceived less humanness in the faces that belonged to their gender outgroup than in the faces that characterised their own gender (t (31) = 2.23, p =.033, d = .39).

Our objective for Study 1 was to replicate this preliminary finding with a larger sample. Furthermore, we systematically explored any effect of age by testing an equal number of 5-year-olds and 6-year-olds. Children’s explicit preference for the groups was also measured as a manipulation check. All of the measures that were administered across both studies are reported in full.

42

Method

Participants

Thirty-two 5-year-olds (16 boys, Mage = 5;7, age range = 5;0-5;11) and 32 6-year-olds (16 boys, Mage = 6;6, age range = 6;0-6;11) were recruited from a local school and a museum to take part in the study. Six more children also participated but were excluded from analysis due to language and hearing difficulties (n = 1), making two or more errors during initial training on the response scale (n = 3, see below), technical error (n = 1) and experimenter error (n = 1). The sample sizes for Studies 1 and 2 were based on previous work with adults using a very similar paradigm (see Hackel et al., 2014) and research examining the development of intergroup bias (e.g., Dunham et al., 2011; Kinzler et al., 2009; Martin et al., 2008). The sample size was chosen in advance and data collection was stopped once the pre-specified sample size was reached.

Materials

Face stimuli. A subset of animacy morphs (N = 10) made available by

Hackel et al. (2014) were used for the training and experimental trials. The morphs were originally developed by combining images of inanimate faces (e.g., dolls, statues) with well-matched human faces, resulting in 11 images that ranged from 0% human to 100% human (see Looser & Wheatley, 2010, for examples of the morph continua). As mentioned above, we carried out a pretest study with adults to identify the most ambiguous morph image for each face identity.

Pretest of face stimuli with adults. We included eight morphs in the test

43 phase. To determine which images along the continuum to use for the test trials, we conducted a pretest study in which we asked 10 adult participants to categorise every morph image (N = 88) as either a doll or a human. Participants were also asked to rate how confident they were with their decision on a 5-point scale (1 = Extremely

uncertain, 3 = Fairly certain, 5 = Extremely certain). The morph that received the

most contradictory set of responses (approx. 50% doll and 50% human) signified the perceptual mid-point for each face identity. Certainty ratings were recorded in order to discriminate between morphs of the same face continuum that received an identical number of contradictory doll-human responses. In this case (n = 4), the image with the lower average certainty rating was selected. Note that the subjective mid-point of the eight morph continua was rarely compatible with the image generated at 50% increment. Five face identities had their highest ambiguity rating at increments greater than 50%, while the remaining continua were perceived to be most ambiguous at the physical mid-point (n = 2) or slightly lower (n = 1). The morphs (see Figure 1) were presented in the approximate dimensions of a life-size face in a central location on a black background using a Lenovo ThinkPad Intel Core i5 laptop.

Figure 1. The face morph stimuli (4 female, 4 male) used in the test trials for Studies